Asymmetric price transmission in the
Japanese seafood market
YUTARO SAKAI, TORU NAKAJIMA1, TAKAHIRO MATSUI2, AND NOBUYUKI YAGI*1
1
Graduate School of Agricultural and Life Sciences, The University of Tokyo, Yayoi ,
Bunkyo, Tokyo, 113-8657, 2Mie University, Graduate School of Bioresources, Tsu City,
Mie Prefecture, 514-8507, Japan,
*Tel: 81-3-5841-5599. Fax: 81-3-5841-5599. Email: yagi@fs.a.u-tokyo.ac.jp
0
Summary
Japanese seafood distribution consists of three levels of market, namely
producer’s, wholesale and retailers market. This system has developed to
manage a large quantity of diverse seafood products rapidly, but its function
is said to have been imperfect. The study mainly focused on six fish species
namely Japanese sardine, horse mackerel, Japanese flying squid, skipjack
tuna, Pacific saury, and red seabream. This study analysed the price
formation function of the Japanese seafood distribution for the six major fish
species through asymmetric price transmission (APT) analysis. The results
indicated that there were several APTs in the market most likely due to a
combined effect of perishable nature of fisheries products, existence of
substitute goods, and imperfect completion of the market.
Key words: asymmetric price transmission (APT), market competition, price formation,
seafood distribution, Japan.
1
Table of Contents
1. Introduction ................................................................................................ 3
2. Materials and Methods .............................................................................. 3
3. Data............................................................................................................. 5
4. Unit root test .............................................................................................. 7
5. Analysis model.......................................................................................... 10
6. Results ...................................................................................................... 12
7. Discussion ................................................................................................. 18
Reference ......................................................................................................... 22
2
1. Introduction
Japanese seafood products are traditionally distributed through three
markets namely landing site producers’ market, consuming site wholesale
markets, and retailers. This distribution system has developed to manage a
large quantity of diverse seafood products rapidly. Recently, however, several
shortcomings of the traditional distribution system are pointed out.
Lou(2009)1) pointed out restructuring the markets is urgent and fishermen,
wholesales and brokers must leave their traditional distribution system and
create a new business.
While most of such arguments are based on qualitative observations, a few
quantitative studies also exist. Ariji(2006)2) analysed market integration of
snow crab between landing markets and wholesales markets in Kyoto
prefecture. The focus of this study was the extent of the integration of prices
in different markets and it didn’t refer to function of distribution in detail.
Therefore the quantitative knowledge about the function of Japanese seafood
distribution system is still insufficient. The purpose of this study is to
analyze the function of Japanese seafood distribution from the view of price
formation.
2. Materials and Methods
2.1 Asymmetric price transmission (APT)
In the economically perfect competitive market, when the exogenous shock
against the market equilibrium happens, it is said the price is adjusted
instantly and new equilibrium occurs. In the real market, however, price will
never be adjusted instantly. The speed of adjustment can be different from
price increasing scenario to price decreasing scenario. Thus, Asymmetric
price transmission (APT) can be detected under such situation. Positive APT
is usually defined as a set of reactions according to which any price
3
movement that squeezes the margin is transmitted more rapidly than the
equivalent movement that stretches the margin. Conversely, APT is negative
when price movements that stretch the margin are transmitted more rapidly
than movements that squeeze it. Gonzales et al.(2003)3) detected the APT in
the distribution of wild cod and farmed salmon and made a comment that the
reason of APT is due to market power of retailer. Jaffry(2005)4) detected the
APT in the distribution of hakes in France and made the same comment as
Gonzales referring to retailer’s market power. Matsui et al.(2011)5) analysed
Japanese blue fin tuna market by separating the target period in ten years
and discussed that entities having the market power shifted from upstream
to downstream by tuna market structure change.
Some other studies argued that a cause of the asymmetry price transmission
is due to a property of the adjustment expense with the corporate activities.
Ward (1982)6) pointed out that the price of perishable agricultural products
cannot be raised as easily as decreasing the price. Jordi and Ramon(2008)7)
analyzed three stages of distribution called a producer’s, consumer’s and
retailer’s market of 12 seafood products in Spain and detected APTs with
many cases. They discussed that the cause is not due to imperfect competitive
market but due to the existence of substitute products in the lower side of the
distribution channel.
In this study, we used time series model to detect a presence of asymmetry
price transmission. The focus was mainly on (i) landing site producers’
markets, (ii) consuming site wholesale markets and (iii) retailers. Price
transmission between (i) and (ii), as well as (ii) and (iii) were examined using
the price data of same fish species. We also conducted a time wise comparison
using the price date of same fish species on or before 1993 and thereafter. The
dividing year was chosen based on Demura(2002)8) which reported that in
1994 the number of consumers who purchase seafood from supermarkets
exceeded to those who purchase seafood in traditional fish mongers.
4
2.2 Data
Six popular fish species in the Japanese market were chosen for the analysis.
These are Japanese sardine, horse mackerel, Japanese flying squid, skipjack
tuna, Pacific saury, and red seabream. Although tuna is popular fish in Japan,
it is not used for the analysis because of data discontinuation i.e. frequent
changes on categorisation of tuna (i.e., coverage of tuna species) in market
statistics during this past 20 years.
The price data are collected in three different stages in the market value
chain. These are (i) producer’s markets in landing ports, (ii) wholesale
markets in consumer’s cities and (iii) retailers. Fish prices in (i) and (ii) were
taken from published data from the Ministry of Agriculture, Forestry and
Fisheries, Japan9) and retailer’s prices were taken from statistics of Ministry
of Internal Affairs and Communications, Japan10).
For some species, prices for fresh and frozen forms of fish are available. If this
is the case, weighted averages of the prices for fresh and frozen fish were
used for the analysis. Fish which were once frozen and then thawed are
categorized as fresh in the statistics. Wholesale market prices are the
weighted mean of the prices of the six main wholesale markets in Japan
(Tokyo, Yokohama, Nagoya, Osaka, Kyoto, Kobe). The retail price was
calculated by the simple average of the retail prices in Tokyo, Yokohama,
Nagoya, Osaka, Kyoto, and in Kobe.
Monthly data are used for analysis. Data are taken from January 1976 to
December 2009 for Japanese sardine, horse mackerel, Pacific saury, and
common squid. The data used for red seabream are from January 1976 to
December 2006. For skipjack tuna, the data from January 1983 to December
2009 are used. Time series of data used for Seabream and skipjack tuna are
slightly shorter than those in other species, since statistical categories have
been changed for these two species and only the data which have clear
5
consistency in category are used for the analysis of the two species. No data
were used for "the frozen saury in Kyoto”, "frozen common squid in
Yokohama", and "frozen horse mackerel after September 2005”. This is
because several products in different processing stages were found under the
same name in the statistics of these particular fish. In addition, since
skipjack tuna statistics in retail level were not available in winter seasons
(from November to February), retail prices for skipjack tuna for these months
are supplemented using Karman filter method11).
Table 1 indicates some details of these data. Red seabream generally have
higher prices than other fish, and the average prices at wholesale markets
are more than 1,000 yen/kg and the average price at retail markets is above
3,000 yen/kg. Skipjack tuna has the second highest average price in retail
markets with more than 2,000 yen/kg. Prices for other four species are
cheaper. Generally, price of each fish species increases from wholesale to
retail markets.
Table 1 Descriptive statistics
Variable (Yen/kg)
Mean
S.D.
Min
Max
Producer
73.53
112.15
9.00
822.00
Wholesale
248.40
126.50
92.00
666.64
Retail
643.61
280.39
281.67
1326.67
Producer
289.60
183.34
34.00
1030.00
Wholesale
514.68
84.94
292.46
866.88
Retail
1598.48
239.68
915.00
2356.67
Producer
315.93
138.75
99.19
867.95
Wholesale
467.28
101.69
262.87
790.97
Retail
1027.85
186.52
666.67
1576.67
174.09
41.92
91.42
310.09
Sardine
Horse mackerel
Japanese flying squid
Skipjack tuna
Producer
6
Wholesale
585.87
185.40
244.00
1232.74
Retail
2346.15
362.28
1436.67
3183.33
Producer
187.61
211.54
29.00
2027.00
Wholesale
390.54
155.91
157.23
1106.70
Retail
1015.48
282.42
566.67
2606.67
Producer
1287.68
337.35
630.00
2181.00
Wholesale
1429.60
414.36
715.62
2399.00
Retail
3044.80
433.35
2150.00
3751.67
Pacific saury
Red seabream
Producer: Producer’s wholesale market price, Wholesale: Wholesale’s wholesale market
price, Retail: Retail market price. Producer’s price and wholesale’s price were obtained
from Annual Statistics on Marketing of Fishery Products. Retail price was obtained
from Annual Report on the Retail Price Survey.
2.3 Unit root test
In time series analyses, there is a situation where a high coefficient and a
significant t-value can be observed while no true relationship in fact exist12).
In other words, spurious regression is more likely to be detected when time
series data was used. To avoid picking up spurious correlations, unit root test
and co-integration test are to be employed. As for unit root test, Augmented
Dickey-Fuller test13) and Kwiatkovski-Phillips-Schmidt-Shin (KPSS) test14)
were conducted before starting price transmission analysis. The results are
shown in Table 2. The null hypothesis, which assumes variables are
non-stationary, can be dismissed for almost all price data. For many of the
first difference, the null hypothesis was not rejected by 1% significance. This
result suggests that every price data series are non-stationary but becomes
stationary by first difference. Thus, co-integration test, which examines
whether true relations exist between non-stationary data15), needs to be
performed to confirm the estimated result have true relationship.
7
Table 2 Unit root test results
(a) Augmented Dickey-Fuller test
Level
Variable
Trend
Const.
Trend
First difference
None
Trend
Const.
Trend
None
Sardine
Producer
-2.02
-1.29
-0.68
-21.34***
-21.36***
-21.39***
Wholesale
-1.59
-1.11
0.45
-23.27***
-23.30***
-23.32***
Retail
-0.87
-1.10
1.08
-14.08***
-14.09***
-14.09***
-2.84
-1.78
-1.30
-20.94***
-20.97***
-20.99***
Wholesale
-4.22***
-2.45
-0.31
-22.99***
-23.01***
-23.03***
Retail
-5.21***
-2.38
-0.14
-25.23***
-25.23***
-25.26***
Horse mackerel
Producer
Japanese flying squid
Producer
-3.13
-1.90
-1.14
-18.29***
-18.31***
-18.33***
Wholesale
-3.32*
-2.39
-0.75
-20.09***
-20.11***
-20.13***
Retail
-3.16*
-1.74
-0.45
-16.21***
-16.22***
-16.24***
-3.64**
-3.57***
-0.75
-20.80***
-20.84***
-20.87***
Wholesale
-2.62
-2.54
-0.12
-16.40***
-16.42***
-16.45***
Retail
-1.76
-1.96
0.07
-14.12***
-14.14***
-14.16***
-2.62
-2.63*
-1.36
-27.22***
-27.26***
-27.29***
-3.65**
-3.59***
-1.33
-23.77***
-23.79***
-23.82***
-4.20***
-4.16***
-0.81
-24.94***
-24.97***
-25.00***
Producer
-3.10
-0.82
-1.34
-22.77***
-22.80***
-22.83***
Wholesale
-2.98
-1.50
-1.08
-23.93***
-23.96***
-23.98***
-4.16***
0.05
-1.22
-20.33***
-20.35***
-20.37***
Skipjack tuna
Producer
Pacific saury
Producer
Wholesale
Retail
Red seabream
Retail
* 10% significance level. ** 5% significance level. *** 1% significance level.
Producer: Producer’s wholesale market price, Wholesale: Wholesale’s wholesale market
price, Retail: Retail market price. Optimal lag length was determined by modified
Akaike Information Criteria.
8
(b) Kwiatkovski-Phillips-Schmidt-Shin test
Level
Variable
Trend
Const.
First difference
Const.
Trend
Const.
Const.
Sardine
Producer
0.43***
1.92***
0.05
0.11
Wholesale
0.23***
2.36***
0.12
0.15
Retail
0.25***
2.42***
0.12
0.16
0.38***
1.88***
0.09
0.09
0.11
1.38***
0.11
0.24
0.20**
1.07***
0.13*
0.33
0.27***
1.93***
0.07
0.07
0.11
1.38***
0.06
0.08
0.23***
1.52***
0.06
0.14
0.10
0.37*
0.27***
0.28
0.21**
0.36*
0.10
0.11
0.50***
0.60**
0.06
0.06
Producer
0.10
0.25
0.21**
0.26
Wholesale
0.12*
0.31
0.05
0.07
Retail
0.12*
0.13
0.04
0.11
Producer
0.24***
2.13***
0.12*
0.13
Wholesale
0.20**
2.11***
0.06
0.06
0.46***
2.13***
0.08
0.17
Horse mackerel
Producer
Wholesale
Retail
Japanese flying squid
Producer
Wholesale
Retail
Skipjack tuna
Producer
Wholesale
Retail
Pacific saury
Red seabream
Retail
* 10% significance level. ** 5% significance level. *** 1% significance level.
Producer: Producer’s wholesale market price, Wholesale: Wholesale’s wholesale market
price, Retail: Retail market price. Optimal Newey-West bandwidth was determined
using Bartlett kernel. The null hypothesis is that the time series is stationary; while the
alternative hypothesis is that the series is non-stationary. Note that the hypothesis
setting of KPSS test is different from that of ADF test.
9
2.4 Analysis model
When the price series have unit root, the long-term relationship can be
described as follows:
𝑌𝑡 = α + β𝑋𝑡 + 𝜇𝑡
(1)
𝑌𝑡 is market price of the lower side (e.g., closer position to consumers) of the
fish distribution channel, and 𝑋𝑡 is market price of the upper side (e.g.,
closer position to producers) of the channel. When the time series data on the
two prices are non-stationary, the estimation using Ordinary Least-Squares
method (OLS) could result in spurious regression. According to Engle and
Granger(1987)15), when a residual 𝜇𝑡 provided by estimation is stationary,
the estimated result is not spurious but represents true relations.
The estimation in this analysis therefore should be conducted using equation
(2):
∆𝜇𝑡 = 𝜌1 𝜇𝑡−1 + ∑𝑇𝑖=1 𝛾𝑖 ∆𝜇𝑡−𝑖 + 𝜀𝑡
(2)
𝑇 is the length of the lag and 𝜀𝑡 is white noise disturbance term. Adjustment
of the series of residuals from previous period to the current period will be
expressed in 𝜌1 𝜇𝑡−1 and it means that APT price adjustment process in the
market is assumed to be symmetric against any exogenous shocks.
If a price adjustment process is asymmetric, the Threshold Autoregressive
(TAR) model, proposed by Enders and Granger(1998)14), needs to be used to
avoid estimation bias. In this model, co-integration test is conducted by
equation (3) instead of equation (2):
∆𝜇𝑡 = 𝐼𝑡 𝜌1 𝜇𝑡−1 + (1 − 𝐼𝑡 )𝜌2 𝜇𝑡−1 + ∑𝑇𝑖=1 𝛾𝑖 ∆𝜇𝑡−𝑖 + 𝜀𝑡
(3)
10
1　if　 t 1  
It  
0　if　 t 1  
(4)
𝐼𝑡 is called Heaviside indicator and the definition is shown in equation (4).
Suppose that 𝜇𝑡−1 is bigger than the threshold τ, adjustment of this period is
shown in 𝜌1 𝜇𝑡−1 . If 𝜇𝑡−1 is smaller than the threshold τ, adjustment of this
period is shown in 𝜌2 𝜇𝑡−1 . When 𝜌1 = 𝜌2 is hold, it means that the
adjustment process is symmetric and, in this case, equation (3) can also
represent equation (2)
Under equation (3), the necessary and sufficient condition for the stationary
residual series can be shown in the following equation (5)16):
(5)
𝜌1 < 0 , 𝜌2 < 0, (1 + 𝜌1 )(1 + 𝜌2 ) < 1
In equation (5), a null hypothesis ( 𝜌1 = 𝜌2 = 0 ) under no existence of
co-integration can be examined through F-test modified by Engle and
Siklos(2001)17) and Wane et al.(2004)18). If there is co-integration, normal
F-test can be used to examine the null hypothesis on symmetric adjustment
process (𝜌1 = 𝜌2 ).
In addition to the TAR model, Momentum-Threshold Autoregressive
(M-TAR) model have been used. It is sometimes suggested that M-TAR model
is more suitable than TAR model for the analysis on perishable products such
as foods19) M-TAR model can be expressed as follows:
∆𝜇𝑡 = 𝑀𝑡 𝜌1 𝜇𝑡−1 + (1 − 𝑀𝑡 )𝜌2 𝜇𝑡−1 + ∑𝑇𝑖=1 𝛾𝑖 ∆𝜇𝑡−𝑖 + 𝜀𝑡
1　if　 t 1  
Mt  
0　if　 t 1  
(6)
(7)
In this study, estimations are made using both TAR and M-TRA model and
the better model will be chosen in light of Akaike information criterion (AIC).
Many previous studies assumed the threshold is 0 a priori, while the validity
of this assumption is not fully examined. In this study, the estimation of the
11
threshold itself also was performed by the procedures suggested in Chan
(1993)20). Specifically, we start the estimation process using Equation (1) by
normal OLS. Then, residuals obtained here are listed, and the upper and the
lower 15% of them are removed. The remaining 70% of the residual values
could be a candidate of the threshold. These candidates are used for the
estimation by TAR and M-TAR model. The candidate which minimizes the
sum of the squared residuals is chosen. The threshold chosen by this
procedure becomes the super-consistent indicator.
3. Results
The estimated results of six species by both TAR and M-TAR model are
shown in Table 3 and Table 4.
Table 3 TAR model result
(a) Producer’s to wholesale’s wholesale market
First period (-1993)
Sardine
Horse
Japanese
Skipjack
Pacific
Red
mackerel
flying squid
tuna
saury
seabream
𝜌1
-0.08
-0.27***
-0.37***
-0.38***
-0.09**
-0.27***
𝜌2
-0.20
-0.18***
-0.25***
-0.29***
-0.39***
-0.05
Lags
2
0
0
1
0
3
𝜏
-0.18
-0.10
-0.10
0.13
-0.35
-0.08
Hypotheses test:
Φ
7.32**
32.10***
47.18***
16.64***
37.60***
7.57**
Asym.
2.53
1.16
1.49
0.60
12.81***
6.59**
Q(6)
6.23
6.41
3.63
5.24
9.86
9.37
AIC
-127.09
-233.71
-226.09
-41.39
-56.18
-247.81
Diagnostic:
Second period (1994-)
Sardine
𝜌1
-0.58***
Horse
Japanese
Skipjack
Pacific
Red
mackerel
flying squid
tuna
saury
seabream
-0.72***
-0.48***
-0.83***
-0.23***
-0.34***
12
𝜌2
-0.49***
-0.68***
-0.41***
-0.68***
-0.39***
-0.27***
Lags
0
1
0
4
0
0
𝜏
0.13
0.10
0.09
0.29
-0.28
0.08
Hypotheses test:
Φ
69.85***
44.96***
55.71***
41.59***
14.03***
30.48***
Asym.
0.41
0.07
0.35
2.12
3.97**
0.30
Q(6)
9.14
5.59
4.67
4.48
14.69
4.05
AIC
-116.57
-193.35
-148.17
-18.66
-3.57
-188.65
Diagnostic:
* 10% significance level. ** 5% significance level. *** 1% significance level.
Φ is the F statistics testing the null hypothesis of 𝜌1 = 𝜌2 = 0. Asym. is the F statistics
testing the null hypothesis of 𝜌1 = 𝜌2. Q(6) is the Ljung-Box Q statistics testing the null
hypothesis of no serial correlation up to six orders. AIC is the Akaike Information
Criteria.
(b) Wholesale’s wholesale to retail market
First period (-1993)
Sardine
Horse
Japanese
Skipjack
Pacific
Red
mackerel
flying squid
tuna
saury
seabream
𝜌1
-0.34***
-0.23***
-0.43***
-0.20**
-0.11**
-0.45***
𝜌2
-0.23***
-0.11*
-0.21***
-0.37***
-0.19**
-0.19***
Lags
0
4
3
2
1
0
𝜏
0.07
-0.11
0.12
0.08
-0.13
-0.03
Hypotheses test:
Φ
36.04***
6.20*
24.11***
7.81**
8.27**
50.66***
Asym.
1.37
1.56
7.94***
1.75
0.98
6.83***
Q(6)
10.11
0.66
1.56
4.41
10.23
4.51
AIC
-241.46
-241.51
-279.66
-142.83
-238.55
-444.81
Diagnostic:
Second period (1994-)
Sardine
Horse
Japanese
Skipjack
Pacific
Red
mackerel
flying squid
tuna
saury
seabream
𝜌1
-0.62***
-0.41***
-0.66***
-0.31***
-0.49***
-0.25***
𝜌2
-0.28***
-0.18**
-0.37***
-0.14
-0.79***
-0.07
Lags
1
3
3
0
1
1
13
𝜏
-0.07
0.03
-0.10
-0.11
-0.10
0.06
Hypotheses test:
Φ
22.78***
7.86**
41.15***
28.96***
26.63***
8.92**
Asym.
6.91***
3.48*
8.41***
3.33*
3.55*
5.18**
Q(6)
6.52
1.86
4.58
7.67
4.11
2.10
AIC
-232.21
-313.29
-238.90
-245.69
-151.51
-331.61
Diagnostic:
* 10% significance level. ** 5% significance level. *** 1% significance level.
Φ is the F statistics testing the null hypothesis of 𝜌1 = 𝜌2 = 0. Asym. is the F statistics
testing the null hypothesis of 𝜌1 = 𝜌2. Q(6) is the Ljung-Box Q statistics testing the null
hypothesis of no serial correlation up to six orders. AIC is the Akaike Information
Criteria.
The estimated results of TAR model can be summarized as follows:
Ljung-Box statistic values are insignificant for all cases thus, the residuals
do not have serial correlation for all the 24 combinations of six spices, two
observation periods, and two combinations of market stages in value chain.
Lag numbers are 0 to 4 for each fish. Φ statistic values are significant in 17
cases out of the 24 combinations by 1% significant level, while by 10% level
all are significant. Thus, all price series have co-integration. APTs are
identified between producer’s and consumer’s wholesale market for saury
and red seabream in early period (-1993) and saury in the late period (1994-).
APTs are also identified between consumer’s wholesale market and retailers
for common squid and red seabream in early period (-1993) and for all the six
species in the late period (1994-).
14
Table 4 M-TAR model result
(a) Producer’s to wholesale’s wholesale market
First period (-1993)
Sardine
Horse
Japanese
Skipjack
Pacific
Red
mackerel
flying squid
tuna
saury
seabream
𝜌1
-0.09**
-0.29***
-0.35***
-0.39***
-0.80***
-0.18***
𝜌2
-0.30***
-0.20***
-0.25**
-0.20*
-0.17***
-0.06
Lags
2
0
0
1
0
3
𝜏
-0.11
-0.02
-0.09
-0.10
0.26
-0.02
Hypotheses test:
Φ
8.81**
39.03***
45.99***
17.69***
34.75***
5.09
Asym.
5.35**
1.04
0.51
2.27
9.80***
1.84
Q(6)
5.40
6.15
4.18
6.02
6.25
10.22
AIC
-128.50
-236.58
-224.31
-42.23
-54.44
-245.43
Diagnostic:
Second period (1994-)
Sardine
Horse
Japanese
Skipjack
Pacific
Red
mackerel
flying squid
tuna
saury
seabream
𝜌1
-0.74***
-0.90***
-0.49***
-0.95***
-0.25***
-0.43***
𝜌2
-0.42***
-0.67***
-0.37***
-0.66***
-0.39***
-0.21***
Lags
0
1
0
4
0
0
𝜏
0.13
0.10
-0.04
0.17
-0.09
0.05
Hypotheses test:
Φ
74.55***
46.42***
55.91***
45.05***
9.88**
30.46***
Asym.
5.55**
2.04
0.82
6.92***
1.62
3.40*
Q(6)
8.08
4.89
4.26
4.06
11.84
5.92
AIC
-117.97
-194.35
-147.64
-21.08
-3.38
-189.14
Diagnostic:
* 10% significance level. ** 5% significance level. *** 1% significance level.
Φ is the F statistics testing the null hypothesis of 𝜌1 = 𝜌2 = 0. Asym. is the F statistics
testing the null hypothesis of 𝜌1 = 𝜌2. Q(6) is the Ljung-Box Q statistics testing the null
hypothesis of no serial correlation up to six orders. AIC is the Akaike Information
Criteria.
15
(b) Wholesale’s wholesale to retail market
First period (-1993)
Sardine
Horse
Japanese
Skipjack
Pacific
Red
mackerel
flying squid
tuna
saury
seabream
𝜌1
-0.35***
-0.30***
-0.24***
-0.23***
-0.11**
-0.47***
𝜌2
-0.13***
-0.11*
-0.66***
-0.71***
-0.55***
-0.30***
Lags
0
4
2
2
2
0
𝜏
-0.04
0.05
-0.09
-0.12
-0.09
0.03
Hypotheses test:
Φ
40.56***
7.38*
21.62***
10.41***
13.48***
47.03***
Asym.
4.78**
3.80*
5.41**
6.44**
12.66***
1.92
Q(6)
8.52
0.67
3.23
5.07
8.14
5.42
AIC
-242.94
-242.65
-279.64
-145.17
-244.60
-441.31
Diagnostic:
Second period (1994-)
Sardine
Horse
Japanese
Skipjack
Pacific
Red
mackerel
flying squid
tuna
saury
seabream
𝜌1
-0.73***
-0.69***
-0.50***
-0.62***
-0.46***
-0.10***
𝜌2
-0.44***
-0.38***
-0.61***
-0.20***
-1.32***
-0.26***
Lags
0
0
3
1
1
1
𝜏
0.04
0.05
-0.02
0.08
-0.07
-0.05
Hypotheses test:
Φ
76.91***
61.09***
36.25***
18.56***
37.23***
7.29*
Asym.
4.79**
4.41**
1.35
10.18***
20.36***
2.17
Q(6)
10.25
10.63
4.32
7.14
2.23
1.49
AIC
-229.86
-311.86
-235.37
-247.55
-159.54
-330.11
Diagnostic:
* 10% significance level. ** 5% significance level. *** 1% significance level.
Φ is the F statistics testing the null hypothesis of 𝜌1 = 𝜌2 = 0. Asym. is the F statistics
testing the null hypothesis of 𝜌1 = 𝜌2 . Q(6) is the Ljung-Box Q statistics testing the null
hypothesis of no serial correlation up to six orders. AIC is the Akaike Information
Criteria.
16
The estimated results of M-TAR model can be summarized as follows:
Ljung-Box statistic values are also insignificant in all the cases and there are
no serial correlations in the residuals. The number of lag is also 0 to 4 for
each fish. Φ
statistic values are significant in 19 cases out of 24
combinations by 1% significant level, while 23 cases are significant by 10%
level. Therefore price series have co-integration relationship in most of the
cases except for one case for red seabream. APTs are identified between
producer’s and consumer’s wholesale market for saury and Japanese sardine
in early period (-1993) and saury, red seabream, and skipjack tuna in the late
period (1994-). APTs are also identified between consumer’s wholesale
market and retailers for five species except for red seabream in early period
(-1993) and for four species except for seabream and common squid in the late
period (1994-).
Some inconformity exists between the results of APT by TAR and M-TAR
models. Thus, the model, which has the smaller AIC, was chosen for each fish
species. Table 5 shows the results sorted out by this procedure. APTs are
identified between producer’s and consumer’s wholesale market for saury
(positive APT), sardine (positive APT), and red seabream (negative APT) in
early period (-1993) and saury (positive), red seabream (negative), skipjack
tuna (negative), and sardine (negative) in the late period (1994-). APTs are
also identified between consumer’s wholesale market and retailers for all
species in early period (-1993) and in the late period (1994-).
Positive and negative APT are defined as follows according to Meyer and
Cramon-Taubadel(2004)21). When upper market price is increased or
decreased, the positive APT is defined that the reaction of lower market price
is faster in increased price scenario than in decreased price scenario. The
negative APT definition is that the reaction in the lower market price is
faster when decreased than when increased. When positive APT exists, the
upper sellers can enjoy the sales margin while the lower buyers can enjoy the
17
margin in the negative APT situation. According to the definition in the TAR
and M-TAR model, when |𝜌1 | < |𝜌2 | the positive APT exists and when |𝜌1 | >
|𝜌2 | the negative APT exists
In this study, while other 5 species have negative APT in most of the cases,
saury always had positive APT with any period and distribution stages.
Throughout the two different periods, while most fish species show the same
continuing result of positive or negative APTs, Japanese sardine and skipjack
tuna changed from positive to negative or no-APT to negative. No fish species
exist which changed APTs from negative to positive.
Table 5 Summary of the results
Period
Producer
Sardine
Horse
mackerel
Japanese
flying
squid
Skipjack
Pacific
Red
tuna
saury
seabream
First
Pos**
－
－
－
Pos***
Neg**
Wholesale
Second
Neg**
－
－
Neg***
Pos**
Neg*
Wholesale
First
Neg**
Neg*
Neg***
Pos**
Pos***
Neg***
Second
Neg***
Neg*
Neg***
Neg***
Pos***
Neg**
to
to
Retail
* 10% significance level. ** 5% significance level. *** 1% significance level.
Producer: Producer’s wholesale market, Wholesale: Wholesale’s wholesale market,
Retail: Retail market, First: The period before 1994, Second: The period in 1994 and
thereafter. Pos and Neg shows the positive and negative asymmetry, respectively.
4. Discussion
In previous literatures, three possible reasons of APT have been identified.
These are perishable nature of fisheries products, existence of substitute
goods, and imperfect competition of the market.
First, if perishable nature of fish products is the cause of APTs, negative
asymmetry can be observed. This is because the sales sides are usually afraid
18
of the increased stockpile of unsold items for such products. This may lead to
the situation where the buyers have more power than the sellers’ when
negotiating the prices and negative APT can be observed. This is particularly
true in Japanese market where the popular form of the consumption of fish is
sashimi or sushi (i.e., raw fish) and consumers are extremely sensitive on the
freshness of the fish products. In this study, negative APTs are in fact found
in most of the species including sardine, horse mackerel, and red seabream.
Their usual form of the distribution is fresh, rather than frozen. Perishable
nature of the products can be one reason for the negative APTs for these
species.
Second, if the existence of substitute products is the cause of APTs, negative
asymmetry can be observed. The reason is closely related to that of
perishable products: i.e., the sales sides may prefer to sell the products at
cheaper prices to compete with the substitutes rather than allowing to
excessive accumulation of stockpiles in perishable products. This may also
lead to the buyers’ power when negotiating the prices. In this study, skipjack
tuna has changed from positive (or no-APT) to negative APTs after 1993.
Large portion of skipjack tuna are distributed in frozen condition and no
evidence exist to support increased perishability of skipjack tuna in the
recent years. The negative APTs for this species are, therefore, attributable
to other element. We consider that the cause of negative APTs for skipjack
tuna can be explained by the existence of substitute goods. Skipjack tuna and
other tuna species (including bluefin tuna, southern bluefin tuna, bigeye
tuna, yellowfin tuna, albacore tuna, and swordfish) have a strong substitute
relationship in the Japanese consumer market22), and consumption of such
substitutable tuna species has increased after early 1990s.
Third, if the imperfect competition in the market is the cause of an APT,
positive asymmetry can be observed when sellers are strong while negative
asymmetry can be observed when buyers are strong. In this study, most of the
19
fish have negative APTs. This result suggests the existence of market powers
in the buyers’ side. The number of species which have negative APTs was
increased in the latter half of the observation period. These outcomes
correspond to the fact that traditional fishmongers were rather quickly
replaced by supermarket around 1994. Because the number of supermarkets
is less than the number of sellers at the wholesale markets, supermarkets
most likely have stronger market powers than their predecessors (i.e.,
traditional fish retailers). Negative APTs for most of the species other than
saury can be explained by the possible imperfect competition in the market
by the increasing dominance of powerful supermarkets, along with the
perishable nature and existence of substitute goods.
Fourth, positive asymmetry for saury suggests that market power exists for
sellers of this particular fish. We have conducted interviewing survey to
further investigate the reason for this peculiar outcome. According to
managers and workers in supermarkets and wholesale markets, saury is
exceptionally important fish for sales promotion activities for the
supermarket as a whole in recent years. Japanese fresh saury is used as a
consumer attraction item and it is usually placed in the top page of
supermarket leaflet. While supermarket usually provides service for its
customers in filleting or gutting of many fish species, saury is usually
purchased by consumers even though the fish is round (i.e., not filleted nor
gutted). In other words, saury is easy merchant item to sell from the
viewpoint of supermarket. Thus, there is a pulling power from supermarket
for saury even under the situation where retail margins are minimal. These
finding from interviewing are in conformity with the outcome of this study on
APTs.
If this market condition continue to exist in Japan, upstream side (fishermen,
for instance) of the seafood (except for saury) distribution chain would be
forced to accept their long standing weak position vis-a-vis their downstream
counterparts in the value chain.
20
One option to reconcile this situation for fishermen would be to simplify the
fish distribution channel. Current system, which is based on the traditional
system, is comprised of five players. These are (i) sales representative at
producers markets in landing sites, (ii) buyers at producers markets who
tranship fish to the city wholesale markets, (iii) wholesalers at city wholesale
markets, (iv) buyers at city wholesale markets, and (v) retailers. Simplifying
the value chain may be achieved through some technological development
such as introduction of e-commerce combined with efficient collecting and
transporting methods of fish products. This restructuring process may
strengthen the connection among sales representatives of fishermen by
e-commerce and putting the fishermen in better position in the distribution
channel.
Another option to improve the negative APT due to perishability of the raw
products may be a policy that promotes sales of frozen fishes. Retailers
usually defrost fishes when displaying in their shops because defrosted
products are more preferable for and familiar to consumers. If fishes are sold
frozen and the consumers defrost them at home, freshness of the products is
maintained and, more importantly, the problem of negative APT due to
perishability could be mitigated.
It should be noted, however, that although we discussed possible causes of
APTs separately, in reality the market would be affected by combined effects
of perishable nature of fisheries products, existence of substitute goods, and
imperfect market competition. The relative magnitude of these elements was
not assessed in this study, and future works are awaited on this aspect.
21
Reference
1) Lou X. Change of fresh seafood product distribution system and building of supply
chain (in Japanese). Journal of Food System Research 2009; 16-2: 59-73.
2) Ariji M. Quantitative analysis of fisheries economic (in Japanese). Seizando
publication, Tokyo. 2006.
3) Gonzales F, Guillotreau P, Grel L, Simioni M. Asymmetry of Price Transmission
within the French Value Chain of Seafood Products. Economics Working Paper Archive
(Toulouse) 2003; 49: 1-22.
4) Jaffry S. Asymmetric Price Transmission: A Case 1 Study of the French Hake Value
Chain. Marine Resource Economics 19; 4: 511-523.
5) Matsui T, Nakajima T, Sakai Y, Yagi N. Structural Change and Price
Transmission in Japanese Marine Products Market: The Case of Bluefin Tuna
(in Japanese). Kokusai Gyogyo Kenkyu 2011; 10: 1-13.
6) Ward RW. Asymmetry in Retail, Wholesale, and Shipping Point Pricing for Fresh
Vegetables. American Journal of Agricultural Economics 1982; 64-2: 205-212.
7) Jordi G, Ramon F. Analysis of the Price Transmission along the Spanish Market
Chain for different Seafood Products. Conference Papers The XVIIIth Annual EAFE
Conference 9th to 11th July 2007; Reykjavik, Iceland.
8) Demura M. Changes in seafood distribution and response from production sites (in
Japanese). Norin Kinyu 2002; 672: 38-51.
9) Seafood Product Distribution Statics. the Ministry of Agriculture, Forestry and
Fisheries. Tokyo. 1976～2009.
10) Retail price statistics. Ministry of Internal Affairs and Communications. Tokyo.
1976～2009.
11) Jones R. Maximum Likelihood Fitting of ARMA Models to Time Series With
Missing Observations. Technometrics 1980; 22: 389-395.
12) Granger CW, Newbold P. Spurious regressions in econometrics. Journal of
Econometrics 1987; 2: 111-120.
13) Dickey D, Fuller W. Distribution of the estimators for autoregressive time-series
with a unit root. Journal of the American Statistical Association 1979; 74: 427-431.
14) Enders W, Granger CW. Unit-Root Tests and Asymmetric Adjustment with an
Example Using the Term Structure of Interest Rates. Journal of Business and Economic
Statistics 1998; 16-3: 304-311.
15) Engle RF, Granger CW. Co-integration and Error Correction: Representation,
Estimation, and Testing. Econometrica 1987; 55-2: 251-276.
22
16) Bai J, Perron P. Computation and Analysis of Multiple Structural Change Models.
Journal of Applied Econometrics 2003; 18-1: 1-22.
17) Enders W, Siklos PL. Cointegration and Threshold Adjustment. Journal of
Business and Economic Statistics 2001; 19-2: 166-176.
18) Wane A, Gilbert S, Dibooglu S. Critical Values of the Empirical F-Distribution for
Threshold Autoregressive and Momentum Threshold Autoregressive Models. Southern
Illinois University Department of Economics Faculty Discussion Papers 2004; 23: 1-16.
19) Frey G, Manera M. Econometric Models of Asymmetric Price Transmission.
Journal of Economic Surveys 2007; 21-2: 349-415.
20) Chan KS. Consistency and Limiting Distribution of the Least-Squares Estimator
of a Threshold Autoregressive Model. The Annals of Statistics 1993; 21-1: 520-533.
21) Meyer J, Cramon-Taubadel SV. Asymmetric Price Transmission: A Survey. Journal
of Agricultural Economics 2004; 55-3: 581-611.
22) Sakai Y, Yagi N, Ariji M, Takahara A, Kurokura H. Substitute and complement
relations among various fish species in the Japanese market: implications for fishery
resource management. Fisheries Science 2009; 75-5:1079-1087.
23